From: Johannes Weiner <hannes@cmpxchg.org>
To: Michal Hocko <mhocko@suse.cz>
Cc: Andrew Morton <akpm@linux-foundation.org>,
Hugh Dickins <hughd@google.com>, Tejun Heo <tj@kernel.org>,
Vladimir Davydov <vdavydov@parallels.com>,
linux-mm@kvack.org, cgroups@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: Re: [patch 12/13] mm: memcontrol: rewrite charge API
Date: Mon, 14 Jul 2014 13:13:24 -0400 [thread overview]
Message-ID: <20140714171324.GQ29639@cmpxchg.org> (raw)
In-Reply-To: <20140714150446.GD30713@dhcp22.suse.cz>
On Mon, Jul 14, 2014 at 05:04:46PM +0200, Michal Hocko wrote:
> Hi,
> I've finally manage to untagle myself from internal stuff...
>
> On Wed 18-06-14 16:40:44, Johannes Weiner wrote:
> > The memcg charge API charges pages before they are rmapped - i.e. have
> > an actual "type" - and so every callsite needs its own set of charge
> > and uncharge functions to know what type is being operated on. Worse,
> > uncharge has to happen from a context that is still type-specific,
> > rather than at the end of the page's lifetime with exclusive access,
> > and so requires a lot of synchronization.
> >
> > Rewrite the charge API to provide a generic set of try_charge(),
> > commit_charge() and cancel_charge() transaction operations, much like
> > what's currently done for swap-in:
> >
> > mem_cgroup_try_charge() attempts to reserve a charge, reclaiming
> > pages from the memcg if necessary.
> >
> > mem_cgroup_commit_charge() commits the page to the charge once it
> > has a valid page->mapping and PageAnon() reliably tells the type.
> >
> > mem_cgroup_cancel_charge() aborts the transaction.
> >
> > This reduces the charge API and enables subsequent patches to
> > drastically simplify uncharging.
> >
> > As pages need to be committed after rmap is established but before
> > they are added to the LRU, page_add_new_anon_rmap() must stop doing
> > LRU additions again. Revive lru_cache_add_active_or_unevictable().
>
> I think it would make more sense to do
> lru_cache_add_active_or_unevictable in a separate patch for easier
> review. Too late, though...
>
> Few comments bellow
> > Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
>
> The patch looks correct but the code is quite tricky so I hope I didn't
> miss anything.
>
> Acked-by: Michal Hocko <mhocko@suse.cz>
Thanks!
> > @@ -54,28 +54,11 @@ struct mem_cgroup_reclaim_cookie {
> > };
> >
> > #ifdef CONFIG_MEMCG
> > -/*
> > - * All "charge" functions with gfp_mask should use GFP_KERNEL or
> > - * (gfp_mask & GFP_RECLAIM_MASK). In current implementatin, memcg doesn't
> > - * alloc memory but reclaims memory from all available zones. So, "where I want
> > - * memory from" bits of gfp_mask has no meaning. So any bits of that field is
> > - * available but adding a rule is better. charge functions' gfp_mask should
> > - * be set to GFP_KERNEL or gfp_mask & GFP_RECLAIM_MASK for avoiding ambiguous
> > - * codes.
> > - * (Of course, if memcg does memory allocation in future, GFP_KERNEL is sane.)
> > - */
>
> I think we should slightly modify the comment but the primary idea
> should stay there. What about the following?
> /*
> * Although memcg charge functions do not allocate any memory they are
> * still getting GFP mask to control the reclaim process (therefore
> * gfp_mask & GFP_RECLAIM_MASK is expected).
> * GFP_KERNEL should be used for the general charge path without any
> * constraints for the reclaim
> * __GFP_WAIT should be cleared for atomic contexts
> * __GFP_NORETRY should be set for charges which might fail rather than
> * spend too much time reclaiming
> * __GFP_NOFAIL should be set for charges which cannot fail.
> */
What *is* the primary idea here?
Taking any kind of gfp mask and interpreting the bits that pertain to
you is done in a lot of places already, and there really is no need to
duplicate the documentation and risk it getting stale and misleading.
> > @@ -948,6 +951,7 @@ static int do_huge_pmd_wp_page_fallback(struct mm_struct *mm,
> > struct page *page,
> > unsigned long haddr)
> > {
> > + struct mem_cgroup *memcg;
> > spinlock_t *ptl;
> > pgtable_t pgtable;
> > pmd_t _pmd;
> > @@ -968,20 +972,21 @@ static int do_huge_pmd_wp_page_fallback(struct mm_struct *mm,
> > __GFP_OTHER_NODE,
> > vma, address, page_to_nid(page));
> > if (unlikely(!pages[i] ||
> > - mem_cgroup_charge_anon(pages[i], mm,
> > - GFP_KERNEL))) {
> > + mem_cgroup_try_charge(pages[i], mm, GFP_KERNEL,
> > + &memcg))) {
> > if (pages[i])
> > put_page(pages[i]);
> > - mem_cgroup_uncharge_start();
> > while (--i >= 0) {
> > - mem_cgroup_uncharge_page(pages[i]);
> > + memcg = (void *)page_private(pages[i]);
>
> Hmm, OK the memcg couldn't go away even if mm owner has left it because
> the charge is already there and the page is not on LRU so the
> mem_cgroup_css_free will wait until we uncharge it or put to LRU.
Yep, res_counter charges have always pinned the memcg. We already
used this exact protocol and relied on the same lifetime rules for
swapin charging.
> > +/**
> > + * mem_cgroup_commit_charge - commit a page charge
> > + * @page: page to charge
> > + * @memcg: memcg to charge the page to
> > + * @lrucare: page might be on LRU already
> > + *
> > + * Finalize a charge transaction started by mem_cgroup_try_charge(),
> > + * after page->mapping has been set up. This must happen atomically
> > + * as part of the page instantiation, i.e. under the page table lock
> > + * for anonymous pages, under the page lock for page and swap cache.
> > + *
> > + * In addition, the page must not be on the LRU during the commit, to
> > + * prevent racing with task migration. If it might be, use @lrucare.
> > + *
> > + * Use mem_cgroup_cancel_charge() to cancel the transaction instead.
> > + */
> > +void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
> > + bool lrucare)
>
> I think we should be explicit that this is only required for LRU pages.
> kmem doesn't have to finalize the transaction.
The function itself only applies to user/LRU pages. kmem has its own
separate API for charge/commit/cancel/uncharge.
next prev parent reply other threads:[~2014-07-14 17:13 UTC|newest]
Thread overview: 61+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-06-18 20:40 [patch 00/13] mm: memcontrol: naturalize charge lifetime v4 Johannes Weiner
2014-06-18 20:40 ` [patch 01/13] mm: memcontrol: fold mem_cgroup_do_charge() Johannes Weiner
2014-06-18 20:40 ` [patch 02/13] mm: memcontrol: rearrange charging fast path Johannes Weiner
2014-06-18 20:40 ` [patch 03/13] mm: memcontrol: reclaim at least once for __GFP_NORETRY Johannes Weiner
2014-06-18 20:40 ` [patch 04/13] mm: huge_memory: use GFP_TRANSHUGE when charging huge pages Johannes Weiner
2014-06-18 20:40 ` [patch 05/13] mm: memcontrol: retry reclaim for oom-disabled and __GFP_NOFAIL charges Johannes Weiner
2014-06-18 20:40 ` [patch 06/13] mm: memcontrol: remove explicit OOM parameter in charge path Johannes Weiner
2014-06-18 20:40 ` [patch 07/13] mm: memcontrol: simplify move precharge function Johannes Weiner
2014-06-18 20:40 ` [patch 08/13] mm: memcontrol: catch root bypass in move precharge Johannes Weiner
2014-06-18 20:40 ` [patch 09/13] mm: memcontrol: use root_mem_cgroup res_counter Johannes Weiner
2014-06-18 20:40 ` [patch 10/13] mm: memcontrol: remove ordering between pc->mem_cgroup and PageCgroupUsed Johannes Weiner
2014-06-18 20:40 ` [patch 11/13] mm: memcontrol: do not acquire page_cgroup lock for kmem pages Johannes Weiner
2014-06-18 20:40 ` [patch 12/13] mm: memcontrol: rewrite charge API Johannes Weiner
2014-06-23 6:15 ` Uwe Kleine-König
2014-06-23 9:30 ` Michal Hocko
2014-06-23 9:42 ` Uwe Kleine-König
2014-07-14 15:04 ` Michal Hocko
2014-07-14 17:13 ` Johannes Weiner [this message]
2014-07-14 18:43 ` Michal Hocko
2014-06-18 20:40 ` [patch 13/13] mm: memcontrol: rewrite uncharge API Johannes Weiner
2014-06-20 16:36 ` [PATCH -mm] memcg: mem_cgroup_charge_statistics needs preempt_disable Michal Hocko
2014-06-23 4:16 ` Johannes Weiner
2014-06-21 0:34 ` [patch 13/13] mm: memcontrol: rewrite uncharge API Sasha Levin
2014-06-21 0:56 ` Andrew Morton
2014-06-21 1:03 ` Sasha Levin
2014-07-15 8:25 ` Michal Hocko
2014-07-15 12:19 ` Michal Hocko
2014-07-18 7:12 ` Michal Hocko
2014-07-18 14:45 ` Johannes Weiner
2014-07-18 15:12 ` Miklos Szeredi
2014-07-19 17:39 ` Johannes Weiner
2014-07-22 15:08 ` Michal Hocko
2014-07-22 15:44 ` Miklos Szeredi
2014-07-23 14:38 ` Michal Hocko
2014-07-23 15:06 ` Johannes Weiner
2014-07-23 15:19 ` Michal Hocko
2014-07-23 15:36 ` Johannes Weiner
2014-07-23 18:08 ` Miklos Szeredi
2014-07-23 21:02 ` Johannes Weiner
2014-07-24 8:46 ` Michal Hocko
2014-07-24 9:02 ` Michal Hocko
2014-07-25 15:26 ` Johannes Weiner
2014-07-25 15:43 ` Michal Hocko
2014-07-25 17:34 ` Johannes Weiner
2014-07-15 14:23 ` Michal Hocko
2014-07-15 15:09 ` Johannes Weiner
2014-07-15 15:18 ` Michal Hocko
2014-07-15 15:46 ` Johannes Weiner
2014-07-15 15:56 ` Michal Hocko
2014-07-15 15:55 ` Naoya Horiguchi
2014-07-15 16:07 ` Michal Hocko
2014-07-15 17:34 ` Johannes Weiner
2014-07-15 18:21 ` Michal Hocko
2014-07-15 18:43 ` Naoya Horiguchi
2014-07-15 19:04 ` Johannes Weiner
2014-07-15 20:49 ` Naoya Horiguchi
2014-07-15 21:48 ` Johannes Weiner
2014-07-16 7:55 ` Michal Hocko
2014-07-16 13:30 ` Naoya Horiguchi
2014-07-16 14:14 ` Johannes Weiner
2014-07-16 14:57 ` Naoya Horiguchi
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20140714171324.GQ29639@cmpxchg.org \
--to=hannes@cmpxchg.org \
--cc=akpm@linux-foundation.org \
--cc=cgroups@vger.kernel.org \
--cc=hughd@google.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mhocko@suse.cz \
--cc=tj@kernel.org \
--cc=vdavydov@parallels.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).